CBH statistics for the Provisional Review

Curtis Seaman, Yoo-Jeong Noh, Steve Miller and Dan Lindsey

CIRA/Colorado State University12/27/2013

Cloud Base Height Evaluation

CBH algorithm for liquid clouds:

Red variables come from upstream retrievals

LWC is pre-defined average value based on cloud type; cloud type comes from upstream retrieval

The cloud base height for liquid clouds is defined at right. Cloud base height definition for ice clouds is similar, except the average ice water content is temperature dependent.

CBH requires upstream retrievals of cloud top height (CTH), cloud optical depth (t), effective particle size (re) and cloud type, which is used to determine the LWC value to use.

Errors in CBH are directly proportional to errors in each of these values. Issues in upstream retrievals directly impact CBH retrieval.

CloudSat/VIIRS CBH validation• Collocated VIIRS pixels/CloudSat profiles where the two satellites are within 15

min define “match-up periods” – Match-up periods occur for ~4.5 hours every 2-3 days

• Results shown here include all match-up periods from September 2012 and September 2013

• CloudSat profiles where precipitation is detected are excluded– CloudSat cannot detect CBH in precipitation

• VIIRS retrievals of clouds with CTH below 1 km AGL are excluded – CloudSat cannot detect these clouds due to ground clutter

• Collocated pixel/profile pairs where CloudSat and VIIRS both detected cloud and retrieved CTH and CBH without error are “valid matchup points”

• Daytime CBH retrievals only– CloudSat does not operate at night due to battery issues

Match-up locations Sept. 2012 Match-up locations Sept. 2013

CBH evaluation – known issues

2012 2013

Known issue with CTH retrieval (boundary layer liquid cloud tops too high due to temperature inversion) has been resolved since Beta review

CBH difficult to retrieve in cases of cloud overlap

Known issue with CTH retrieval: cirrus cloud tops too low due to CTT

Known issue with CBH retrieval: cirrus cloud too thick due to IWC parameterization

Gray shading represents vertical extent of clouds from CloudSat cloud mask. Colored areas represent vertical extent of clouds from VIIRS CTH and CBH retrievals, sorted by VIIRS cloud type retrieval (from COP).

CBH evaluation – issues caused by cloud type retrieval

Failure to detect cloud overlap

Inconsistent cloud type and CTH; thin clouds identified as “opaque ice”

Gray shading represents vertical extent of clouds from CloudSat cloud mask. Colored areas represent vertical extent of clouds from VIIRS CTH and CBH retrievals, sorted by VIIRS cloud type retrieval (from COP). Black line denotes surface elevation.

CBH retrieval performance is highly dependent on accuracy of upstream retrievals.

High clouds classified as “water”

“All Clouds” vs. “Within Spec”

• The VIIRS CBH algorithm has been evaluated for two groups:– All clouds observed by CloudSat and VIIRS– Only those clouds where the VIIRS CTH retrieval is within the

error specifications (aka “Within Spec”)• Error specifications: CTH must be within 1 km if the COT is greater

than 1, or within 2 km if the COT is less than 1

• Thus, “All Clouds” results show the general performance of the CBH retrieval, “Within Spec” results show the performance of the CBH retrieval when the CTH retrieval is accurate– CBH accuracy is very closely related to CTH accuracy

• CBH is within the error specifications if CBH error is less than 2 km

All Clouds CBH statistics

2012 2013

Low cloud bias removed since Beta2012 2013

Matchup periods examined 6 9

Valid matchup points 36,314 56,655

Percentage of valid points where CTH is “within spec” 40.4% 37.6%

Percentage of valid points where CBH error < 2 km 50.6% 44.6%

Some reduced performance due to more observed high (ice) clouds where CTH is less accurate

2012 2013

All Clouds CBH statistics

Negative errors indicate CloudSat CBH was lower than VIIRS CBH(VIIRS biased high relative to CloudSat)

All Clouds CBH statistics

All Clouds Opaque Ice Cirrus Water Mixed-phase Overlap

Percentage of valid points (%) 100 5.5 36.6 18.9 14.4 24.6

Average Error (km) 0.8 -1.1 1.7 0.9 -0.2 0.6

Median Error (km) 0.6 -1.0 2.2 0.0 -0.3 1.2

Standard Deviation (km) 3.6 3.4 3.5 2.9 2.5 4.2

RMSE (km) 3.6 3.6 3.9 3.0 2.5 4.3

Percentage within 250 m (%) 1.6 0.9 1.6 4.3 1.9 1.4

R-squared correlation (-) 0.188 0.030 0.093 0.124 0.066 0.000

2012

2013

All Clouds Opaque Ice Cirrus Water Mixed-phase Overlap

Percentage of valid points (%) 100 0.6 29.5 25.7 19.2 25.0

Average Error (km) 1.1 1.4 2.2 0.4 0.2 0.5

Median Error (km) 0.5 1.1 2.2 -0.6 -0.2 0.4

Standard Deviation (km) 3.3 3.2 3.1 2.9 2.5 3.8

RMSE (km) 3.4 3.5 3.8 3.0 2.5 4.0

Percentage within 250 m (%) 1.2 0.1 2.4 2.3 1.5 1.6

R-squared correlation (-) 0.212 0.092 0.136 0.087 0.030 0.007

Green values indicate improvement compared to 2012Red values indicate reduced performance compared to 2012

Black values indicate no change compared to 2012

Overall, the average error (bias) has been slightly reduced compared to 2012.

Small increases in the error standard deviation and RMSE are primarily due to relatively poor performance of the CTH retrieval for ice clouds.

Ice clouds are a larger proportion of the total clouds observed in 2013 compared to 2012.

CBH Statistics when CTH is “within spec”

2012 2013

All Cloud Types All Cloud Types

CBH Statistics when CTH is “within spec”

2012 2013

Negative errors indicate CloudSat CBH was lower than VIIRS CBH(VIIRS biased high relative to CloudSat)

CBH statistics when CTH is “within spec”

All Clouds Opaque Ice Cirrus Water Mixed-phase Overlap

Percentage of valid points (%) 100 4.2 28.6 31.1 19.3 16.6

Average Error (km) 0.2 0.5 1.0 -0.2 -0.7 0.8

Median Error (km) -0.1 0.2 0.9 -0.2 -0.4 0.5

Standard Deviation (km) 2.1 2.4 2.7 0.6 1.5 2.8

RMSE (km) 2.1 2.4 2.8 0.7 1.6 2.9

Percentage within 250 m (%) 22.9 10.9 7.3 44.4 26.5 8.1

R-squared correlation (-) 0.595 0.190 0.208 0.814 0.224 0.181

2012

2013

All Clouds Opaque Ice Cirrus Water Mixed-phase Overlap

Percentage of valid points (%) 100 0.1 32.1 32.8 17.8 16.7

Average Error (km) 0.5 1.7 1.6 -0.6 -0.7 1.6

Median Error (km) -0.1 1.9 1.2 -0.6 -0.5 1.4

Standard Deviation (km) 2.2 2.9 2.7 0.7 1.5 2.2

RMSE (km) 2.3 3.4 3.1 0.9 1.7 2.7

Percentage within 250 m (%) 14.6 4.8 8.1 22.5 17.2 9.4

R-squared correlation (-) 0.551 0.199 0.149 0.712 0.191 0.400

Green values indicate improvement compared to 2012Red values indicate reduced performance compared to 2012

Black values indicate no change compared to 2012

When the CTH retrieval is within the error specifications, the CBH retrieval performs better.

In most cases, errors have been reduced, the error standard deviation and RMSE have been reduced, the percentage of “correct” retrievals has increased and the correlations have increased compared to 2012. The exception is “overlap” clouds where the VIIRS retrievals are expected to perform poorly.

CBH Summary

• Based on a comparison with CloudSat– CBH performs much better when CTH is accurate– CTH appears to be more accurate for liquid-topped clouds than for

ice clouds• A known error with the CTH retrieval for low liquid clouds has been fixed,

which improved CBH• Improving CTH for ice clouds will improve CBH

– IWC parameterization produces Cirrus cloud CBH that is much too low

– Inconsistent or incorrect cloud type retrievals negatively impact CBH

• Optically thin “opaque ice” clouds are present in the data set, as are liquid “water” clouds above 10 km MSL – these issues still need to be resolved

Backup Slides

All Clouds

Within Spec

CBH performance – Opaque IceSeptember 2013

All Clouds

Within Spec

CBH performance – CirrusSeptember 2013

All Clouds

Within Spec

All Clouds

Within Spec

CBH performance – WaterSeptember 2013

All Clouds

Within Spec

All Clouds

Within Spec

CBH performance – Mixed-phaseSeptember 2013

All Clouds

Within Spec

All Clouds

Within Spec

CBH performance – OverlapSeptember 2013

All Clouds

Within Spec

All Clouds

Within Spec